Press one key and say the whole thing. It lands as finished text wherever your cursor already is: the AI chat, the email, the message, about three times faster than your hands manage it1. It runs on your machine, not our servers2, so there is no account, no word cap, and nothing you say leaves the device.
Fastspoken is a voice dictation app for Android and Windows, free on both. You press one key, say the whole thing, and it lands as finished text in any app you can type in, about two seconds after your last word. The same key works on the phone and at the desk, so the thought goes wherever you are when you have it. The free plan runs entirely on your own device: no internet needed, no weekly word cap, no trial clock, and not a single word sent to a server. There is no server bill behind your words, which is why free stays free.
Typing makes you edit while you are still thinking. You compress the idea down to what your hands can carry, and the first thing cut is always the same: the tangent, the caveat, the actually, more like this. Those cuts are the context. A model reads your false start as a boundary and your correction as a standard, so the mess is doing work. Speaking keeps all of it. Say it long and unedited and the answer comes back sharper, because you handed over the whole shape of the thought instead of the tidied summary. Ramble in, precision out.
There is nothing to learn and nothing new to open. Fastspoken is a keyboard, so it turns up wherever your cursor already blinks, on the phone and at the desk, the same key on both. Hold it in ChatGPT, in Gmail, in a client message, in a CRM field, and your words arrive with the punctuation already in place. No window to switch to, no transcript to copy across, no place for the idea to go cold.
Whangārei | kōrero | Ngāpuhi | PTY LTD | Tarei
Names, places, te reo, shorthand. Teach a spelling once and it lands right everywhere, free, on your device.
No user counts and no star ratings, because we have not earned them yet. Four figures describe the whole product, so they are printed at the size they deserve.
words of what you said, sent anywhere3 speech becomes text on the device itself
from your last word to finished in the app you were already in
faster than your hands the whole thought, not the short version
The typing is free because it runs on your machine. That is the same reason nothing you say leaves it.
The fourth number, and why it is what it is
Free forever on your device. Not a trial, not a tease, not a countdown. The plan most people never need to leave.
The catch, honestly: your machine does the work, so your words cost us nothing to type and there is no bill for us to pass on. We charge only when we do run a server for you. That is Pro: free types exactly what you said, Pro finishes it, so the ramble you spoke into a reply lands as the email you meant. Shaping words like that takes a model bigger than a phone can hold, and that is the only reason any of this costs anything.
Most dictation streams your voice to a server, runs it through a model in someone else's building, and streams the text back. Fastspoken skips the trip. The model lives on your own machine, the audio never leaves it, and there is no transcript sitting in a bucket somewhere with your name on it.
Nothing you say leaves your machine.
That is not a policy that could quietly change in an update, and not a promise you have to take on trust. There is no server to send your words to. Nothing can leak from a place it was never sent.
Nothing to set up per app. Fastspoken is a keyboard, so every app that takes text takes your voice, including every AI chat you already talk to all day.
No. On the free plan, speech is decoded on your device and the audio never leaves it. On Pro, the optional cloud polish processes your words and immediately discards them. On every plan our servers never store what you say, and there are no word caps on either plan.
Yes. After a one-time speech-engine download on first run (about 630MB, wifi recommended), the free plan runs entirely on your device, so dictation works with no internet at all. Airplane mode, dead zones, and locked-down networks all behave exactly like home.
Any app you can type in. Fastspoken works through the keyboard layer on Android, so messages, email, documents, browsers, and code editors all receive your words at the cursor, without per-app setup.
Yes, free forever. The free plan runs on your own device, so your words cost us nothing to type and there is no server bill for us to pass on. There is no trial countdown, no weekly word cap, and no account required. We charge only for Pro, where we do run a model for you.
On the hard split of a standard speech benchmark (LibriSpeech test-other, 25 clips), the on-device engine scores 96.4%, matching the large cloud-grade model it replaced at 95.7% while running entirely on your phone. Clean speech scores higher. We name the benchmark rather than quoting a marketing percentage.
Gboard commits each word as it hears it and never goes back. Fastspoken types as you talk too, but it keeps re-decoding in the background as the rest of the sentence arrives, so words it already put down get corrected once the context turns up. Punctuation comes out of the engine rather than out of you saying “comma” and “full stop” aloud. It also learns your personal dictionary, runs on Windows as well as Android, and on the free plan never touches a network.
Android and Windows are both live in early access, downloaded directly from this site while the Play listing is in review. iOS is planned and not built yet. Cross-device dictionary and settings sync is live on the Pro plan, end-to-end encrypted.
By default, nothing is kept anywhere, on any plan. Pro adds an optional history toggle that stores your finished dictations on your device only. It never syncs and never reaches a server.